Run batch sentiment analysis on all entries

Usage

analyze_sentiment(
  data,
  provider,
  config = list(),
  checkpoint = NULL,
  research_focus = "",
  coding_state = NULL,
  audit_log = NULL,
  response_cache = NULL
)

Arguments

data: Standardized tibble with std_text column
provider: AIProvider object
config: Sentiment config section
checkpoint: CheckpointManager (or NULL). On resume after a crash, entries already scored in the step's partial checkpoint are adopted (matched by std_id) and skipped, so only unscored entries are re-sent to the provider.
research_focus: Research focus string
coding_state: ProgressiveCodingState (or NULL). When provided, only processes entries in the analytic sample (those with codes) and includes assigned codes as context for more accurate sentiment scoring.
audit_log: An AuditLog object (from init_audit_log) for recording each sentiment-assignment decision, or NULL to disable audit logging for this step.
response_cache: An optional ResponseCache object (from init_response_cache). When provided, raw API responses for each per-batch sentiment ai_complete_fast() call are written to the cache and referenced from the audit log. Pass NULL to skip raw-response capture.

Value

tibble with sentiment_score, confidence, all_emotions (semicolon-separated), emotion_intensity columns added